AI is doing real work in open source. Answering questions, reviewing PRs, writing patches. Some communities ban it, others label it. Most will accept it because policing AI is exhausting and the tooling is useful.

Here's what bothers me. Everyone argues about allowing AI contributions. Nobody talks about what we lose when humans stop doing the work. AI can write code. But it can't show up to community calls for two years. It can't help someone push their first PR. It can't convince a burned-out maintainer to stay. Leadership isn't a pull request. It's a relationship.

We have a leadership problem. Projects lose maintainers faster than they grow new ones. AI makes it worse by paving over entry-level work that used to get people involved.

The policy landscape is messy. Apache requires disclosure. OpenTelemetry treats AI as a tool. Linux Kernel won't accept patches without a human behind them. These policies reveal how communities define contribution, accountability, and belonging.

My argument is simple. Stop fighting AI. Start investing in what it can't do. Mentoring. Building trust. Growing leaders. That's what's at risk.